TUTORIALS

The tutorial aims to provide a low-cost solution for monitoring the impact of climate change and catchment degradation on river water level patterns. The tutorial will cover the development of a river water level data acquisition system. The main sections will be the realization of a water level sensor node and the integration of a wireless sensor network for data transmission from the sensor node to a network server.

The tutorial will also highlight the integration of KENET cloud services in storing the sensor node data from the network server. The sensor node prototypes to be developed will be based on the Arduino Nano, an ATmega328 based microcontroller. The prototypes will also include: an ultrasonic sensor for data acquisition, an RFM95-868MHZ LoRa module setup for data transmission and a DC power supply with a Lithium Ion battery as the source.

Another highlight of the tutorial will be the explanation of how the data is acquired and the data quality control and validation measures with reference to a sensor node deployed along River Muringato in Nyeri County, Kenya.

The data quality control showcase with involve the utilization of machine learning models in detecting anomalies from the data collected. The validation step will involve the comparison of data collected along River Muringato to cause-datasets such as rainfall. The section will also highlight how prediction models can be utilized once other external datasets such as rainfall are incorporated.

Generally, the tutorial is intended for individuals or organizations interested in developing low-cost river water level monitoring systems to manage water resources and preserve ecosystems and researchers interested in the development of open source solutions.

In this 90-minute interactive session, we will discuss the steps taken to design and build a Low-Cost Raspberry Pi-based Camera Trap and how to integrate wildlife detection models into them. The presentation will have five sections: Introduction to Camera Traps, Hardware Development, Software Development, Wildlife Data and Machine Learning Development. The hardware development will present on the physical parts of a camera trap and how to design it for protection and power performance. The software development will include how to capture images and videos using Python programming. Wildlife data will consist of an appropriate description of the DSAIL-Porini dataset. Machine Learning Development will examine model training and deployment on the Raspberry Pi.

Motivation: Ecologists and conservationists perform surveys by counting animals and monitoring their behaviours. These surveys can be conducted physically or using technology. This information is important for conservationists to understand the health of a conservancy. Nowadays using camera traps to acquire wildlife images is becoming more common within conservancies. Camera traps are able to monitor wildlife passively. The hardware and software of the camera traps are usually designed in such a way as to collect as much wildlife imagery at low power consumption as possible. After a period of time, the data is retrieved from the camera traps and processed in the cloud. Since the information is what is of interest to the conservationist, wildlife detection in the camera traps will reduce the workload, making it computationally cheaper to process the data.